New Research

Cognition and Behavior Loss of the Habenula Intrinsic Neuromodulator Kisspeptin1 Affects Learning in Larval Zebrafish

Joanne Chia,5 Vatsala ء,Ruey-Kuang Cheng,4 ء,Mohini Sengupta,3 ء,Charlotte Lupton,1,2 and Suresh Jesuthasan2,4,6 ء,Thirumalai,3

DOI:http://dx.doi.org/10.1523/ENEURO.0326-16.2017 1Department of Animal and Plant Sciences, University of Sheffield, Sheffield, S10 2TN, UK, 2Institute for Molecular and Cell Biology, 138673, Singapore, 3National Centre for Biological Sciences, Tata Institute of Fundamental Research, Bangalore, 560065, India, 4Lee Kong Chian School of Medicine, Nanyang Technological University, 636921, Singapore, 5National University of Singapore Graduate School for Integrative Sciences and Engineering, 117456, Singapore, and 6Duke-NUS Graduate Medical School, 169857 Singapore

Abstract Learning how to actively avoid a predictable threat involves two steps: recognizing the cue that predicts upcoming punishment and learning a behavioral response that will lead to avoidance. In zebrafish, ventral habenula (vHb) neurons have been proposed to participate in both steps by encoding the expected aversiveness of a stimulus. vHb neurons increase their firing rate as expectation of punishment grows but reduce their activity as avoidance learning occurs. This leads to changes in the activity of raphe neurons, which are downstream of the vHb, during learning. How vHb activity is regulated is not known. Here, we ask whether the neuromodulator Kisspeptin1, which is expressed in the ventral habenula together with its receptor, could be involved. Kiss1 mutants were generated with CRISPR/Cas9 using guide RNAs targeted to the signal sequence. Mutants, which have a stop codon upstream of the active Kisspeptin1 peptide, have a deficiency in learning to avoid a shock that is predicted by light. Electrophysiology indicates that Kisspeptin1 has a concentration-dependent effect on vHb neurons: depolarizing at low concentrations and hyperpo- larizing at high concentrations. Two-photon calcium imaging shows that mutants have reduced raphe response to shock. These data are consistent with the hypothesis that Kisspeptin1 modulates habenula neurons as the learns to cope with a threat. Learning a behavioral strategy to overcome a stressor may thus be accompanied by physiological change in the habenula, mediated by intrinsic . Key words: Calcium imaging; electrophysiology; habenula; intrinsic neuromodulation; mutant; operant learning

Significance Statement Learning to deal with adversity can positively affect one’s ability to cope with challenges in the immediate future. Control thus causes short-term change in the brain. Here, we show that the neuromodulator Kisspeptin1 is involved in learning to avoid a punishment. The expression pattern of this gene and electrophysiological recordings suggest that this molecule may function by modulating the ventral habenula, a region of the brain that mediates fear by regulating release. Kisspeptin1 could thus be a potential player in resilience developed as a result of control, extending previous findings that it can reduce fear.

Introduction On first encountering a threat in a novel environment, When faced with an aversive stimulus, animals respond there may be panic and poorly directed attempts at es- in a manner that is dependent on context and experience. cape. If the animal repeatedly encounters the threat and

Received October 25, 2016; accepted March 30, 2017; First published May 8, Author contributions: CL, MS, RKC, VT, and SJ designed research; CL, MS, 2017. and RKC performed research; CL, MS, RKC, JC, and VT analyzed data; SJ Authors report no conflict of interest. wrote the paper.

May/June 2017, 4(3) e0326-16.2017 1–9 New Research 2 of 9 becomes familiar with a safe escape route, it will be able What is the mechanism of change in vHb activity as to quickly remove itself from danger. Better still, if the learning occurs? Amo et al. (2014) have proposed that animal is able to recognize a cue that reliably predicts excitation is regulated via feedback from serotonergic the impending threat, it will be able to escape before the neurons in the raphe to the entopeduncular nucleus, aversive stimulus is present. This, in essence, is the phe- which is the teleost homolog of the (Turner nomenon of active avoidance. Several theories have been et al., 2016). Whether additional mechanisms are involved proposed to explain the mechanism underlying active is unknown. Here, we examine the possibility that a avoidance. In the two-factor theory, the animal first de- change in habenula neurons accompanies the learning velops a fear of the conditioning stimulus (CS) that is process. In particular, we examine the potential involve- paired with the aversive stimulus, by Pavlovian condition- ment of the neuromodulator Kisspeptin1, which is ex- ing. Cotermination of the CS and the threat [uncondi- pressed in the vHb together with its receptor (Kitahashi tioned stimulus (US)] then drives learning. Expectation has et al., 2009). In zebrafish, two paralogs of kiss1 have been a critical role, and actions that lead to better-than- identified. These have nonoverlapping expression, with predicted outcomes are reinforced (Maia, 2012). The first kiss1 being restricted to the habenula and kiss2 ex- time an aversive stimulus is encountered, the predicted pressed in the and posterior tuberculum outcome would be negative. However, if an escape route (Kitahashi et al., 2009). This allows a specific test of the has been learned, then the predicted outcome becomes role of kiss1 in the habenula using genetics. positive. In , kiss1 is expressed in the hypothalamus How is active avoidance implemented in the brain? One and is well studied in the context of reproduction (Pinilla structure that may be involved is the lateral habenula. As et al., 2012; Clarke et al., 2015). Burst firing of hypotha- first shown in monkeys, unexpected punishment leads to lamic neurons leads to the release of Kisspeptin1 (Kelly increased activity in lateral habenula neurons, whereas an et al., 2013), which causes sustained depolarization of unexpected reward leads to inhibition (Matsumoto and gonadotropin-releasing hormone (GnRH) neurons (Han Hikosaka, 2007, 2009). Direct evidence for an involvement et al., 2005). Kisspeptin1 is also expressed in the hip- of this structure in active avoidance comes from ze- pocampus, where it causes an increase in EPSC ampli- brafish, in which the homolog of the lateral habenula is tude and contributes to increased excitability (Arai, 2009). termed the ventral habenula (vHb; Amo et al., 2010). In the zebrafish habenula, Kisspeptin1 has been proposed Silencing vHb neurons reduces the ability of zebrafish to to depolarize vHb neurons, based on its ability to induce learn active avoidance (Amo et al., 2014). In the naive c-fos expression (Ogawa et al., 2014). However, delivery animal, unexpected punishment leads to phasic activity in of Kisspeptin1 decreases fear, which is inconsistent with vHb neurons (Amo et al., 2014). As the animal learns to evidence that excitation of vHb neurons is aversive (Amo associate a CS with the threat, there is tonic firing in et al., 2014). Moreover, killing Kisspeptin1 receptor–ex- response to the CS in vHb neurons. This causes excita- pressing neurons in ventral habenula neurons mimics the tion of serotonergic neurons in the dorsal raphe. As the effect of Kisspeptin1 delivery (Ogawa et al., 2014), which animal learns to escape, there is decreased tonic firing. would not be expected if Kisspeptin1 causes depolariza- Tonic activity in the vHb has thus been proposed to tion. Rather, these observations suggest that Kisspeptin1 encode aversive reward expectation value, and learning can also cause hyperpolarization. Here, we ask whether active avoidance is associated with a change in vHb Kisspeptin1 could be involved in learning avoidance, by activity in response to the CS (Amo et al., 2014). assessing the ability of a kiss1 mutant to learn and whether Kiss1 has the ability to both depolarize and hy- -Star Graduate Academy, under the ARAP scheme. perpolarize vHb neurons, which would make Kiss1 signalءCL was funded by the A MS was supported by Senior Research Fellowship from Council of Scientific ing a potential mechanism for alteration of vHb neuron and Industrial Research. JC was funded by the National University of Singa- activity during avoidance learning. pore Graduate School. This work was supported by core funding from Institute for Molecular and Cell Biology and a Lee Kong Chian School of Medicine, Nanyang Technological University MOE Start-Up Grant to SJ; core funding Materials and Methods from National Centre for Biological Sciences; Intermediate Fellowship from Animals Wellcome Trust-DBT India Alliance; and grant funding from the Department of Biotechnology to VT. Experiments were conducted on the nacre strain of *C. Lupton, M. Sengupta, and R.-K. Cheng contributed equally to this work zebrafish, Danio rerio, in accordance with protocols ap- and should be considered joint first authors. proved by the Institutional Animal Care and Use Commit- Acknowledgments: We thank Satoshi Ogawa and Ishwar Parhar (Brain tee. Animals at the stages tested could not be sexed. Research Institute, Monash) for the antibodies to Kisspeptin1 and the kiss- peptin receptor. We also thank Caroline Kibat (IMCB) for performing the antibody label and associated imaging and Samuel James (University of Antibody labeling Oxford) for help with the behavioral assay. Standard methods were used. Briefly, fish were fixed Correspondence should be addressed to either of the following: Suresh overnight at 4°C in 4% paraformaldehyde in PBS. A so- Jesuthasan. E-mail: [email protected]; or Vatsala Thirumalai. E-mail: lution of PBS with 1% bovine serum albumin (Fraction V; [email protected]. DOI:http://dx.doi.org/10.1523/ENEURO.0326-16.2017 Sigma-Aldrich), 1% DMSO, and 0.1% Triton X-100 was Copyright © 2017 Lupton et al. used to permeabilize the tissue and dilute primary anti- This is an open-access article distributed under the terms of the Creative bodies. The antibodies to Kisspeptin1 (Parhar Lab, Mo- Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is nash University Malaysia, PAS 15133/15134; RRID: properly attributed. AB_2490069) and the receptor Kissr1b have been

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 3 of 9

described previously (Servili et al., 2011; Nathan et al., three wild-type samples were used as controls, and each 2015a). Alexa Fluor 488–conjugated goat anti-rabbit an- test sample was run as a replicate. tibodies (Thermo Fisher Scientific, A11078, lot 792513; RRID:AB_10584486) were used at 1:1000 dilution in PBS. Active avoidance conditioning Imaging was conducted using a Zeiss LSM510 confocal Conditioning was conducted using a two-way cham- microscope with a 40ϫ water immersion objective. ber, as described in Cheng and Jesuthasan (2012), with the addition of a separator made of matte black card- Mutagenesis of the kiss1 locus board between the two compartments. Before the start of Guide RNAs to the kiss1 gene were designed using ZiFiT conditioning, fish were allowed to habituate in the cham- Targeter (Sander et al., 2007). The Basic Local Alignment ber for 20 min. The CS, a red LED, was delivered for 8 s, Search Tool (BLAST) was used to test for off-targets, and and the US, a 25-V pulse, was delivered for 100 ms using only those target sites that yielded no identical off-targets a Grass SD9 stimulator. An interval of at least 6 min was were used. The chosen target sites were integrated into a provided between trials; the next trial commenced only when the fish entered the target area, which is the quad- forward primer (GAAATTAATACGACTCACTATAGGN18GT- TTTAGAGCTAGAAATAGC; Bassett et al., 2013). PCR was rant near the side wall. The entire training session nor- performed with Phusion High-Fidelity polymerase (Thermo mally lasted 90 min. Fish were aged 4–6 weeks, as Fisher Scientific) and a universal reverse primer that defined younger fish could not learn this type of behavior (Valente the remainder of the single-guide RNA (sgRNA) sequence et al., 2012). Fish were genotyped after the assay by (AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAAC- sequencing. Performance in the assay was determined by GGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAC). a computer (i.e., blind to genotype). PCR products were purified (QIAquick PCR purifica- Statistical analysis tion kit; Qiagen) and 0.1 ␮g was transcribed using the Data structure was assessed using the Shapiro–Wilks MEGAshortscript T7 Transcription Kit (Invitrogen). sgRNAs test, and effect size was measured using Cohen’s d for were purified using ammonium acetate precipitation [2.5 ␮L normally distributed data or with r for non–normally distrib- of 0.1 M EDTA (Promega), 5 ␮Lof5M ammonium acetate uted data. To assess p values, Student’s t test was used for solution (Qiagen), 115 ␮L of 100% ethanol)] and stored in data with a normal distribution; otherwise, Mann–Whitney 1-␮L aliquots at –80°C. U test or Wilcoxon matched-pairs test was applied. The CRISPR/Cas9 expression vector pT3Ts-nls-zCas9- nls (Addgene plasmid #46757; Jao et al., 2013) was lin- Electrophysiology earized using XBaI (New England Biolabs). Cas9 mRNA Whole-cell patch-clamp recordings were performed ␮ was produced by in vitro transcription of 1 g template from vHb neurons in 6–8 days postfertilization (dpf) larvae using the mMessage mMachine T3/T7 kit. Capped, poly- using procedures described in Sengupta and Thirumalai adenylated Cas9 mRNA was made using the Poly(A) kit (2015). Briefly, larvae were anesthetized in 0.01% MS222 (Ambion). The reaction was precipitated using 30 ␮L lith- and pinned onto a Sylgard (Dow Corning) dish using fine ium chloride solution (Ambion). RNA was eluted in 30 ␮L tungsten wire (California Fine Wire). The MS222 was then nuclease-free H2O, aliquoted, and stored at –80°C until replaced with external saline (composition in mM: 134 use. A 1-␮L sample was run on a 1% agarose gel along- NaCl, 2.9 KCl, 1.2 MgCl2, 10 Hepes, 10 Glucose, 2.1 side the RiboRuler High Range ladder (Thermo Fisher CaCl2, 0.01 D-tubocurare, pH 7.8, 290 mOsm), and the Scientific) to check for correct size of product (ϳ3 kb). skin over the head was carefully peeled off to expose the A mixture containing ϳ1 ␮g Cas9 mRNA and 400 ng/␮L brain. The recording chamber was then transferred to of sgRNAs was injected into the animal pole of embryos the rig apparatus. All recordings were done in an awake, from a elavl3:GCaMP6f transgenic line (Cheng et al., in vivo condition. The cells were observed using a 60ϫ 2016) at the one-cell stage. To assess the effects, high- water immersion objective on a compound microscope resolution melt analysis (HRMA) was performed using the (Olympus BX61WI). Cells in the ventral most layer of the MeltDoctor HRM Mastermix [Invitrogen, 5 ␮L MeltDoctor habenula and on the lateral most side were targeted for HRM Mastermix, 1 ␮L diluted DNA (20 ng/␮L; obtained by recordings, as these cells express the Kiss1 receptor. The digesting 24-h embryos in 20 ␮g/␮L proteinase K at 55°C ventral neuropil of the habenula was used as a landmark. for 60 min), 0.3 ␮L forward and reverse primer (10 ␮M), Pipettes of tip diameter 1–1.5 ␮m and resistance of 12–16 ␮ and 3.4 L nuclease-free H2O] under the following con- M⍀ were pulled with thick-walled borosilicate capillaries ditions: 95°C 10 min, 40 cycles of [95°C 15 s, 60°C 1 min], (1.5 mm OD; 0.86 mm ID; Warner Instruments) using a 95°C 10 s, 60°C 1 min, 95°C 15 s, 60°C 15 s on the 7500 Flaming Brown P-97 pipette puller (Sutter Instruments). A Fast Real-Time PCR system (Applied Biosystems). Nu- potassium gluconate–based patch internal solution (com-

cleic acid concentration was measured using a Nanodrop position in mM: 115 K gluconate, 15 KCl, 2 MgCl2,10 ND-1000 (Thermo Fisher Scientific). Concentrations of Hepes, 10 EGTA, 4 MgATP, pH 7.2, 290 mOsm) was used nucleic acids were measured by the absorbance at 260 for all recordings. Whole-cell recordings were acquired nm, and the purity of the sample was quantified by the using Multiclamp 700B amplifier, Digidata 1440A digitizer, ratio of sample absorbance at 260/280 nm. A lower limit of and pCLAMP software (Molecular Devices). The data 1.10 was set as the boundary for an acceptable reading were low-pass filtered at 2 kHz using a Bessel filter and for the 260/280 values because of impurities interfering sampled at 20 kHz at a gain of 1. Membrane potentials with the HRMA analysis. During each HRMA reaction, mentioned were not corrected for liquid junction potential,

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 4 of 9

Figure 1. Generation of mutations in the zebrafish kiss1 gene using CRISPR/Cas9. A, B, Dorsal view of zebrafish larvae, labeled with an antibody to Kisspeptin1 (A; cyan) and the Kisspeptin receptor (B; yellow). These fish are from the Et(SqKR11) transgenic line (Lee et al., 2010) and express red fluorescence in afferents from the entopeduncular nucleus (basal ganglia) that terminate in the neuropil of the ventral habenula. The green arrowhead indicates the habenula. The speckles appear to be nonspecific label. C, A schematic of the kiss1 gene. There are two introns. The signal sequence is shown in blue, and the region containing the active Kisspeptin1 peptide is shown in green. D, Partial sequence of the kiss1 gene. The signal sequence is indicated by the blue bar. The position of the two guide RNAs are indicated by the black bars. E, The sequence of the three mutant alleles, together with predicted translations. The asterisks indicate stop codons. The black bars indicate inserted sequences. F,Akiss1sq1sjϪ/Ϫ fish, following labeling with the Kisspeptin1 antibody. No signal could be detected in the habenula. Scale bar ϭ 25 ␮m. Pa, pallium; rHb, right habenula; lHb, left habenula. Images are single optical sections, with anterior to the left. which was measured to be ϩ8 mV for the potassium and K234 (Isca biochemicals). Events were detected of- gluconate–based internal solution. The following sub- fline using Clampfit (Molecular Devices). Graphs were stances were perfused in the bath: tetrodotoxin (1 ␮M, plotted using Microsoft Excel. Statistical analysis was Alomone labs, Israel), zebrafish K10 (Isca Biochemicals), performed using Matlab.

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 5 of 9

Figure 2. The behavior of kiss1 mutants in an active avoidance assay. A, Schematic of the two-way active avoidance chamber. B, Number of crossovers before shock delivery for mutants (n ϭ 31) and WT siblings (n ϭ 37). The black symbols are average values. C, Number of crossovers in a group of mutants (n ϭ 22) trained over two consecutive days. D, Response to the first shock, as measured by swim speed in the first second immediately after the shock. E, Mean swim speed during the 20-min habituation period. F, Response to the first exposure to the red light. This was measured by the ratio of the mean speed in the 2 s after light onset to the mean speed in the 5 s before light onset. In all cases error bars indicate 95% confidence interval.

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 6 of 9

Table 1. Statistical analysis performed Figure Data structure Type of test Result 2B Normal distribution Shapiro–Wilk test W ϭ 0.969; p ϭ 0.0929 t test p ϭ 0.0024 Effect size (Cohen’s d) d ϭ 0.77 2C Nonnormal distribution Shapiro–Wilk test W ϭ 0.889; p ϭ 0.001 Wilcoxon matched pairs test p ϭ 0.011 Effect size (adjusted Cohen’s d) d ϭ 0.91 2D Normal distribution Shapiro–Wilk test W ϭ 0.966; p ϭ 0.057 t test p ϭ 0.13 2E Normal distribution Shapiro–Wilk test W ϭ 0.986; p ϭ 0.673 t test p ϭ 0.22 2F Nonnormal distribution Shapiro–Wilk test W ϭ 0.829; p ϭ 0.001 Mann–Whitney U test p ϭ 0.98 3C Nonnormal distribution Kruskal–Wallis nonparametric test p ϭ 0.012; p ϭ 0.029 followed by post hoc Tukey–Kramer (10 nM vs. 5 ␮M); method of multiple comparison p ϭ 0.034 (100 nM vs. 5 ␮M) 4B Multilevel model analysis Intracluster correlation 0.46 Chi-squared (df ϭ 1) 5.5971 p ϭ 0.015

Calcium imaging DNA derived from injected embryos indicated that the Zebrafish larvae (6–8 dpf) expressing GCaMP6f under guide RNAs were effective. This was confirmed by se- the elavl3 promoter were anesthetized in mivacurium and quencing: eight of eight injected embryos contained mu- embedded in low-melting-temperature agarose (2% in tations at the target site. Injected siblings were grown to E3:5mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM adulthood, and sequencing of F1 fish indicated a trans- MgSO4) in a glass-bottom dish (Mat Tek). They were mission rate of 100%. Three alleles were obtained: one imaged at 1 Hz on a Nikon two-photon microscope consisted of a 20-bp deletion (kiss1sq1sj; Fig. 1D, E), (A1RMP) on a fixed stage upright microscope using a 25ϫ whereas the other two were insertions of 7 and 8 bp water immersion objective (NA ϭ 1.1). The femtosecond (kiss1sq2sj and kiss1sq3sj); kiss1sq3sj also contained a 1-bp laser (Coherent Vision II) was tuned to 920 nm. The stim- deletion, resulting in the same reading frame as kiss1sq2sj. ulus was delivered using a Grass SD9 stimulator. Fish All alleles gave rise to premature stop codons upstream to used in this experiment were derived from offspring of the Kisspeptin1 peptide (Fig. 1E). No label could be de- sq1sj kiss1 heterozygotes and were genotyped by PCR tected in mutant fish by immunofluorescence with an after imaging. antibody to the C terminus of prepro-Kisspeptin1 (Servili et al., 2011; Fig. 1F), further indicating that the mutants Analysis of calcium imaging data lacked the peptide. Cells in the superior raphe were outlined manually in To test their ability to learn to avoid an aversive stimu- ImageJ to create regions of interest (ROIs). The ratio f/f0 lus, juvenile mutants and wild-type animals (4–6 wks of for all ROIs was obtained, where f0 is the average intensity age) were tested individually in a tank with two compart- in a 15-s period before stimulus onset. The average ratio ments (Fig. 2A), similar to a previously described appara- for each cell in the 15-s period after stimulus onset was tus (Lee et al., 2010; Cheng and Jesuthasan, 2012), but calculated. Estimate of the intracluster correlation and with the addition of a partial separator between the com- multilevel analysis was conducted as described in Aarts partments. Each compartment contained a red light (the et al., (2014), using R software. CS) and electrodes that can deliver an aversive shock. Results The light was turned on for8sinthecompartment As in adult zebrafish (Servili et al., 2011; Nathan et al., containing the fish and coterminated with the shock if the 2015a), Kisspeptin1 is present in the ventral habenula of fish remained in the CS side at the end of the CS presen- larval zebrafish, together with its high-affinity receptor tation. If the fish moved to the non-CS side and stayed Kiss1ra (Onuma and Duan, 2012; Fig. 1A, B). The Kiss- there until the end of the CS presentation, no shock was peptin system is thus expressed in an appropriate manner delivered. Fish were exposed to 10 training trials. The to locally regulate vHb neurons even at an early develop- cross score was defined as the number of trials in which mental stage. If Kisspeptin1 signaling is involved in active an individual fish swam to the non-CS chamber before the avoidance learning, fish that lack Kiss1 should be defi- light was turned off. Kiss1 mutants showed a lower cross cient in this process. To test this, we generated mutations score compared with wild types (Fig. 2B; p ϭ 0.0024, d ϭ in the kiss1 locus using CRISPR/Cas9 (Hruscha et al., 0.77; Table 1). We also trained a separate group of mu- 2013; Hwang et al., 2013). Two guide RNAs were de- tants in the same task over two consecutive days. As signed to the signal peptide region of Kisspeptin1 (Fig. shown in Fig. 2C, a significantly better performance was 1C). These were injected into embryos at the one-cell observed on the second day of training compared with stage, together with mRNA for Cas9. HRMA of genomic the performance on the first day (p ϭ 0.011, d ϭ 0.91).

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 7 of 9

These observations suggest that loss of Kiss1 does not abolish the fish’s learning ability but contributes to sub- optimal performance in learning active avoidance. Kiss1 has been shown to reduce the behavioral response to an aversive stimulus (Ogawa et al., 2014). Thus, mutants may have increased behavioral response to the shock. To test this, we analyzed the swim speed immediately after the onset of the first electrical shock. As shown in Fig. 2D, both mutants and wild type responded to the shock with an increase in swim speed (compare with the swim speed before the shock: Fig. 2F). There was no significant differ- ence between the two groups (p ϭ 0.13). We also asked whether mutant fish have increased anxiety. Anxious fish generally behave differently in a novel environment. We thus compared the swim speed in the 20-min habituation period immediately after fish were introduced to the test tank. As shown in Fig. 2E, no difference was observed between mutant and wild type (p ϭ 0.22). We also assessed the response of the fish to the red LED, as anxious fish may respond to this with a startle (Lee et al., 2010). No difference was seen (Fig. 2F, p ϭ 0.98). Collectively, the data suggest that there was no behavioral difference in the response to the shock, the red light, or the novel environment in the beginning of the task. To determine how Kisspeptin1 affects habenula neu- rons, we performed whole-cell patch-clamp recordings from these cells in 6- to 8-dpf wild-type larvae. We bath- applied zebrafish K-10, a 10–amino acid active peptide of Kisspeptin1. 1 ␮M tetrodotoxin was added to the bath solution to globally block network activity. Cells were recorded in voltage clamp mode and were taken through a series of 500-ms voltage steps (Fig. 3A, lower panel). The cellular response (Fig. 3A, upper panel) was recorded before and after bath application of K-10. The recording was allowed to stabilize for 5 min in normal saline before K-10 was applied. Input resistances and holding currents did not change significantly with application of K-10 (p ϭ 0.486 for holding currents and p ϭ 0.587 for input resis- tances, sign test, n ϭ 40 cells). Difference currents were Figure 3. The effect of Kisspeptin1 on vHb neurons. A, Repre- sentative trace of cellular response (top, black) in a cell to voltage then calculated by subtracting the current values after steps (bottom, gray). These recordings were done in the pres- K-10 application from the one before K-10 application ence of 1 ␮M tetrodotoxin to block network activity. B, Difference ␮ (control). This indicates that 5 M K-10 induced an out- current obtained after bath application of 5 ␮M K-10 (n ϭ 10 ward current at depolarized potentials (Fig. 3B, n ϭ 10 cells). The same protocol as in A was done before and after bath cells from 7-dpf larvae). The mean peak amplitude of this application of K-10 for each cell. The traces obtained after were current at 25 mV was 15.8 Ϯ 8.7 pA. To determine subtracted from the traces obtained before application of the whether the Kisspeptin1-evoked current was dose de- peptide for the same cell. C, Dosage response of the difference current. Superimposed scatter and box plots for the difference pendent, we repeated the experiment with 10 nM, 100 nM, currents induced at 25 mV for four different K-10 concentrations and 1 ␮M K10. Lower concentrations of K-10 application (10 nM, 100 nM,1␮M, and 5 ␮M). Difference currents induced at induced lower amplitudes of the peak current at 25 mV 10 and 100 nM are significantly different from that induced at 5 p Ͻ 0.05, Kruskal–Wallis nonparametric test followed byء) Fig. 3C), with 100 nM inducing an inward current, sug- ␮M) gesting that low and high concentrations of Kisspeptin1 post hoc Tukey–Kramer method of multiple comparison, n ϭ 10 have opposite effects on habenula neurons. This sug- cells for each K10 dose). gests that Kisspeptin1 has concentration-dependent ef- fects on vHb neurons. we performed calcium imaging and measured the re- Kiss1 can increase the number of c-fos–expressing sponse of raphe neurons to shock (Fig. 4). Given the cells in the raphe (Ogawa et al., 2012), and the ability of nested design of the experiment (262 cells in 7 mutants Kiss1 to block response to the alarm substance is medi- and 221 cells in 6 wild-type fish), it is possible that differ- ated by the raphe (Nathan et al., 2015b), which is down- ences may be due to genotype as well as an unknown stream of the vHb. We thus asked whether loss of Kiss1 factor (Aarts et al., 2014). To assess this, we determined would affect activation of the raphe neurons. To find out, the intracluster correlation. The value obtained was 0.456.

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 8 of 9

Figure 4. The effect of shock on raphe neurons in kiss1 mutants and siblings. A, Change in fluorescence of GCaMP6f in neurons in the superior raphe of kiss1sq1sj mutants (blue trace) and wild-type siblings (red trace). 50 V was applied for 1 s, 15 s after the start of imaging, at the point indicated by the arrow. The traces indicate mean value while the shading indicates SEM. B, Average z-score of cells in the 15-s period after onset of shock. Each circle represents a single cell, and cells are arranged according to fish. The black dot indicates mean, and the bars indicate 95% confidence interval.

Multilevel analysis was thus conducted, and this indicated provide one way to resolve this puzzle, which is that that there is a large effect of genotype (Cohen’s d ϭ Kisspeptin1 has a concentration-dependent effect on 1.676). The 95% confidence interval of the effect of ge- ventral habenula neurons: low concentrations lead to notype was 0.241–1.434, i.e., higher than 0, indicating depolarization, and high concentrations lead to hyper- that the effect is significant. Calcium imaging thus sug- polarization. In principle, concentration-dependent gests that loss of Kiss1 reduces activation of raphe neu- modulation of neuronal membrane potential may result rons by an aversive stimulus. from the presence of different receptors with different affinities for the ligand, as has been shown for dopa- Discussion mine (Clemens et al., 2012). Alternatively, this effect In this article, we have tested the possibility that the may be achieved via a single receptor coupled to dif- neuromodulator Kisspeptin1 could participate in active ferent downstream activators with distinct activation avoidance learning in zebrafish. Kisspeptin1 is expressed affinities. only in the habenula of zebrafish (Kitahashi et al., 2009), Although the electrophysiological recordings reported primarily in ventral habenula neurons (Servili et al., 2011). here were performed at the larval stage, which provides The Kisspeptin1 receptor Kiss1ra (Onuma and Duan, optimal conditions for identifying and recording vHb neu- 2012) is expressed in habenula neurons (Servili et al., rons, this property of Kisspeptin1 signaling does not ap- 2011). The reduced ability of kiss1 mutants to learn sug- pear to be restricted to larval zebrafish. Ogawa et al. gests that instrumental learning could involve signaling Ϫ11 within the habenula. Given that loss of Kiss1 did not (2012) reported that 10 mol/g body weight of Kisspep- tin1 increased c-fos expression in the habenula and raphe eliminate learning, other mechanisms must also be in- Ϫ9 volved. These may include a change in input from the of adult zebrafish, whereas a higher concentration of 10 basal ganglia (entopeduncular nucleus) via feedback from mol/g did not. Also, a concentration-dependent effect for the raphe (Amo et al., 2014), as well as other circuits such Kisspeptin was reported in GnRH neurons of the medaka as those in the dorsal pallium (Aoki et al., 2013). (Zhao and Wayne, 2012), with only low concentrations A function for Kisspeptin1 in regulating fear re- causing depolarization. sponses was suggested previously based on the find- The experiments here demonstrate that Kisspeptin1 ing that injection of the peptide into the brain ventricle has the necessary properties to be involved in active of adult zebrafish led to depolarization of habenula avoidance by affecting vHb neurons. First, loss of the neurons, as assessed by c-fos expression, and reduced kiss1 gene affects avoidance learning. Second, Kisspep- expression of innate fear (Ogawa et al., 2014). Surpris- tin1 has physiologic effects on vHb neurons. Third, loss of ingly, destruction of cells containing the Kisspeptin1 kiss1 reduces activation of the raphe, which is down- receptor, including vHb neurons, had the same behav- stream of the vHb. Successful learning may be accompa- ioral effect as administering Kisspeptin1. This raises a nied first by Kisspeptin1-mediated excitation, and then by conundrum: how can stimulating a neuron have the inhibition of habenula neurons. This would increase aver- same effect as killing the neuron? The results here sive expectation while the CS is being associated with

May/June 2017, 4(3) e0326-16.2017 eNeuro.org New Research 9 of 9

US, and reduce aversive expectation once the strategy to off-target effects in zebrafish. Development 140:4982–4987. avoid the US has been learned. What controls the release CrossRef Medline of Kisspeptin1 is unknown at present. Hwang WY, Fu Y, Reyon D, Maeder ML, Tsai SQ, Sander JD, Peterson RT, Yeh J-RJ, Joung JK (2013) Efficient genome editing Aside from ventral habenula neurons, additional targets in zebrafish using a CRISPR-Cas system. Nat Biotechnol31:227– for Kisspeptin1 cannot be ruled out, as this modulator is 229. CrossRef expressed in habenula fibers that innervate the ventral inter- Jao L-E, Wente SR, Chen W (2013) Efficient multiplex biallelic ze- peduncular nucleus (IPN; Servili et al., 2011). Kiss1rb, which brafish genome editing using a CRISPR nuclease system. Proc has lower affinity for Kiss1 compared with Kiss1ra (Onuma Natl Acad SciUSA110:13904–13909. CrossRef and Duan, 2012), is expressed in the IPN (Servili et al., 2011). Katz PS, Getting PA, Frost WN (1994) Dynamic neuromodulation of synaptic strength intrinsic to a central pattern generator circuit. Distinguishing whether Kiss1 acts in the habenula or the IPN 367:729–731. CrossRef Medline will require analysis of receptor mutants. At present, how- Kelly MJ, Zhang C, Qiu J, Rønnekleiv OK (2013) Pacemaking kiss- ever, there is no evidence that the ventral IPN is required for peptin neurons. Exp Physiol 98:1535–1543. CrossRef Medline avoidance learning; instead, the dorsal and intermediate IPN Kitahashi T, Ogawa S, Parhar IS (2009) Cloning and expression of appear to be involved (Agetsuma et al., 2010). All evidence kiss2 in the zebrafish and medaka. Endocrinology 150:821–831. so far is thus consistent with the hypothesis that Kisspeptin1 CrossRef Medline functions as an intrinsic neuromodulator (Katz et al., 1994)of Lee A, Mathuru AS, Teh C, Kibat C, Korzh V, Penney TB, Jesuthasan S (2010) The habenula prevents helpless behavior in larval ze- vHb neurons during avoidance learning. brafish. Curr Biol 20:2211–2216. CrossRef Medline Maia DTV (2012) Avoidance learning. In: Encyclopedia of the Sci- References ences of Learning, 403–406, Boston, MA: Springer US. Aarts E, Verhage M, Veenvliet JV, Dolan CV, van der Sluis S (2014) A Matsumoto M, Hikosaka O (2007) Lateral habenula as a source of solution to dependency: using multilevel analysis to accommodate negative reward signals in neurons. Nature 447:1111– nested data. Nat Neurosci 17:491–496. CrossRef Medline 1115. CrossRef Medline Agetsuma M, Aizawa H, Aoki T, Nakayama R, Takahoko M, Goto M, Matsumoto M, Hikosaka O (2009) Representation of negative moti- Sassa T, Amo R, Shiraki T, Kawakami K, Hosoya T, Higashijima vational value in the primate lateral habenula. Nat Neurosci 12:77– S-I, Okamoto H (2010) The habenula is crucial for experience- 84. CrossRef Medline dependent modification of fear responses in zebrafish. Nat Neu- Nathan FM, Ogawa S, Parhar IS (2015a) Neuronal connectivity be- rosci 13:1354–1356. CrossRef tween habenular glutamate-kisspeptin1 co-expressing neurons Amo R, Fredes F, Kinoshita M, Aoki R, Aizawa H, Agetsuma M, et al. and the raphe 5-HT system. J Neurochem 135:814–829. (2014) The habenulo-raphe serotonergic circuit encodes an aver- Nathan FM, Ogawa S, Parhar IS (2015b) Kisspeptin1 modulates sive expectation value essential for adaptive active avoidance of odorant-evoked fear response via two serotonin receptor subtypes danger. Neuron 84:1034–1048. CrossRef Medline (5-HT1A and 5-HT2) in zebrafish. J Neurochem 133:870–878. Amo R, Aizawa H, Takahoko M, Kobayashi M, Takahashi R, Aoki T, Ogawa S, Nathan FM, Parhar IS (2014) Habenular kisspeptin mod- Okamoto H (2010) Identification of the zebrafish ventral habenula ulates fear in the zebrafish. Proc Natl Acad SciUSA111:3841– as a homolog of the mammalian lateral habenula. J Neurosci 3846. CrossRef Medline 30:1566–1574. CrossRef Medline Ogawa S, Ng KW, Ramadasan PN, Nathan FM, Parhar IS (2012) Ha- Aoki T, Kinoshita M, Aoki R, Agetsuma M, Aizawa H, Yamazaki M, benular Kiss1 neurons modulate the serotonergic system in the brain Takahoko M, Amo R, Arata A, Higashijima S-I, Tsuboi T, Okamoto of zebrafish. Endocrinology 153:2398–2407. CrossRef Medline H (2013) Imaging of neural ensemble for the retrieval of a learned Onuma TA, Duan C (2012) Duplicated Kiss1 receptor genes in ze- behavioral program. Neuron 78:881–894. CrossRef Medline brafish: distinct gene expression patterns, different ligand selec- Arai AC (2009) The role of kisspeptin and GPR54 in the hippocam- tivity, and a novel nuclear isoform with transactivating activity. pus. Peptides 30:16–25. CrossRef Medline FASEB J 26:2941–2950. CrossRef Bassett AR, Tibbit C, Ponting CP, Liu J-L (2013) Highly efficient Pinilla L, Aguilar E, Dieguez C, Millar RP, Tena-Sempere M (2012) targeted mutagenesis of Drosophila with the CRISPR/Cas9 sys- Kisspeptins and reproduction: physiological roles and regulatory tem. Cell Rep 4:220–228. CrossRef Medline mechanisms. Physiol Rev 92:1235–1316. CrossRef Medline Cheng R-K, Krishnan S, Lin Q, Hildebrand DGC, Bianco IH, Kibat C, Sander JD, Zaback P, Joung JK, Voytas DF, Dobbs D (2007) Zinc Jesuthasan S (2016) The drives light-evoked activity in Finger Targeter (ZiFiT): an engineered zinc finger/target site design the habenula of larval zebrafish. bioRxiv. doi: 10.1101/047936. tool. Nucleic Acids Res 35:W599–W605. CrossRef Medline Cheng RK, Jesuthasan S (2012) Automated conditioning in larval Sengupta M, Thirumalai V (2015) AMPA receptor mediated synaptic zebrafish. In: Zebrafish Protocols for Neurobehavioral Research. excitation drives state-dependent bursting in Purkinje neurons of Kalueff AV, Stewart AM, eds., 107–120, Springer, New York. zebrafish larvae. eLife. 2015 Sep 29;4. CrossRef Medline Clarke H, Dhillo WS, Jayasena CN (2015) Comprehensive review on Servili A, Le Page Y, Leprince J, Caraty A, Escobar S, Parhar IS, kisspeptin and its role in reproductive disorders. Endocrinol Metab Seong JY, Vaudry H, Kah O (2011) Organization of two indepen- (Seoul) 30:124–141. CrossRef Medline dent kisspeptin systems derived from evolutionary-ancient kiss Clemens S, Belin-Rauscent A, Simmers J, Combes D (2012) Oppos- genes in the brain of zebrafish. Endocrinology 152:1527–1540. ing modulatory effects of D1- and D2-like receptor activation on a CrossRef Medline spinal central pattern generator. J Neurophysiol 107:2250–2259. Turner KJ, Hawkins TA, Yáñez J, Anadón R, Wilson SW, Folgueira M CrossRef Medline (2016) Afferent connectivity of the zebrafish habenulae. Front Neu- Han S-K, Gottsch ML, Lee KJ, Popa SM, Smith JT, Jakawich SK, ral Circuits 10:3512. CrossRef Clifton DK, Steiner RA, Herbison AE (2005) Activation of Valente A, Huang K-H, Portugues R, Engert F (2012) Ontogeny of gonadotropin-releasing hormone neurons by kisspeptin as a neu- classical and operant learning behaviors in zebrafish. Learn Mem roendocrine switch for the onset of puberty. J Neurosci 25:11349– 19:170–177. CrossRef Medline 11356. CrossRef Medline Zhao Y, Wayne NL (2012) Effects of kisspeptin1 on electrical activity of Hruscha A, Krawitz P, Rechenberg A, Heinrich V, Hecht J, Haass C, an extrahypothalamic population of gonadotropin-releasing hormone Schmid B (2013) Efficient CRISPR/Cas9 genome editing with low neurons in medaka (Oryzias latipes). PLoS ONE 7:e37909. CrossRef

May/June 2017, 4(3) e0326-16.2017 eNeuro.org